Using numerical simulations, we investigate to what extent the pump-probe spectrum can be used as a tool to determine the exciton delocalization length in disordered molecular (J) aggregates. We compare the delocalization length obtained through heuristic arguments from the spectral separation between the bleaching and the one- to two-exciton induced absorption features in this spectrum to the delocalization length obtained from the participation ratio at the J band center. We find that up to a certain saturation length these two delocalization lengths are indeed proportional. In the case of long-pulse two-color pump-probe spectra, the slope of this linear scaling is insensitive to the pump frequency; it is sensitive to the width of the probe pulse. Beyond the saturation length, which is determined by the homogeneous line width of the exciton transitions, both the two-color and the short-pulse pump-probe spectra saturate and no longer can be used as a measure for the delocalization length. The latter applies to the B850 ring of bacterial antenna complexes.